Path: utzoo!attcan!uunet!lll-winken!lll-lcc!ames!pasteur!ucbvax!CAEN.ENGIN.UMICH.EDU!pha
From: pha@CAEN.ENGIN.UMICH.EDU (Paul H. Anderson)
Newsgroups: comp.sys.apollo
Subject: Re: why no native unix??
Message-ID: <4064d40e5.000f088@caen.engin.umich.edu>
Date: 21 Dec 88 20:40:57 GMT
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The Internet
Lines: 65

If computer users decide they want absolute consistency, and that
they want 4.3 BSD, then they should really only be using Vax-780's
through vt100's.  Every other machine I have seen has made non-standard
(where the bsd tape is defined as standard) extensions.  It doesn't
matter whether it is sun, dec, apollo, or whatever.

Yes, there are varying degrees of closeness to bsd4.3.  No matter which
machine you choose that is different from that old vax, it shall be
different.  The vax standard is a standard in a similar sense that
the 370 is a standard - one arch, one assembly language, one calling
convention, one compiler, and so on.

It used to be that people who wanted Unix were, in my opinion, best
off with vax hardware, and second best off with Sun hardware.  Again,
this is if you define unix to be 4.3 BSD (or 4.2 awhile ago).  Milage
varies with other unix variants.

Now, with SR10.1, Apollo Unix is sufficiently close to "real" unix (4.3
BSD) that the argument "but they aren't real Unix!" doesn't fly anymore.

Yes, SR10.1 on the Apollos is not real Unix.  It has significant enchancements
that are absolutely imperative in maintaining a large network.  Don't fool
yourself otherwise.

Our relatively small staff here at CAEN support close to 500
Apollos, each of which is fully backed up once a week, and incrementally
backed up once a day, without bringing down the network or the machine.
Each of our machines transparently, painlessly,
quickly, and reliably access files on each other machine.  We run probably
over one hundred million dollars worth of software on these machines, if
you count standard pricing for software (much of this is in the form of
gifts, or grants, of course).

Apollo still has shortcomings.  CAEN still has shortcomings, in part due
to Apollo, and in part due to local political conditions.  However, both Apollo
and CAEN are struggling to make a network of two to ten thousand nodes
administrable by the same staff size that now maintains these 500 and still
provide all the advantages we enjoy now.

I apologize for being so blunt, but there is no way in hell that anyone
anywhere is going to do that with Suns, MicroVaxen, PC/RTs, NeXt boxes,
or any other box that is running "just Unix."  Yes, there are networks
out there that run custom software that attempt to overcome some of
the problems, but they aren't running "real Unix" by a long shot, either.

The real argument that should be made about Unix is how to properly address
some of these issues, such as maintaining a huge network.  Yeah, everybody
is trying to come up with things and ways of doing this, and some certainly
are suceeding.  However, Apollo is really the only vendor that made
the choice to dump a poorly designed system and replace it with something
better.  Now, everyone is struggling to do the same, and pretty much
paying the same price.  The initial price being paid is lower, yes, but it is
stretched out over a much longer period of transition time.

Yes, the Apollo system still has lots of problems, but getting a unix
system that is as close to possible as that 4.3 BSD tape, and also,
incidentally, scales up to thousands of nodes without requiring thousands
of support people is going to be very, very costly.  Apollo did the
right thing in tossing Unix first, rather than the sun approach of tossing
scalability first.

These opinions are my own and sure as hell don't represent anyone
else's...

Paul Anderson
CAEN Apollo systems programmer